Solid State Communications 127 (2003) 433437 www.elsevier.

com/locate/ssc

Raman line-shape analysis of nano-structural evolution in cation-ordered ZrTiO4-based dielectrics

Young K. Kim, Hyun M. Jang*
Department of Materials Science and Engineering, National Research Laboratory (NRL) for Ferroelectric Phase Transitions, Pohang University of Science and Technology (POSTECH), Pohang 790-784, South Korea Received 18 February 2003; accepted 26 May 2003 by H. Akai

Abstract It is known that the microwave dielectric characteristics of ZrTiO4 are greatly inuenced by processing conditions such as the degree of Sn-modication and the cooling rate after sintering. The effects of these processing variables on the nano-structural characteristics of ZrTiO4-based dielectrics were studied by analyzing the Raman line-shape parameters using the phononconnement model. It was shown that the nano-structural shape of the cation-ordered domains underwent a sequential change from thin-slab form to platelet shape, and nally to spherical shape with increasing Sn-content or cooling rate. q 2003 Elsevier Ltd. All rights reserved.
PACS: 78.30.Ly; 77.84.Dy; 64.70.Rh; 64.60.Cn Keywords: A. Nanostructure; A. Oxides; C. Raman spectroscopy; D. Phonons

1. Introduction Dielectric resonator now becomes the key element in modern microwave integrated devices. Dielectric materials for the resonator should possess high dielectric permittivity 1r ; small temperature coefcient of resonant frequency tcf ; and low dielectric loss tan d [1]. Zirconium titanate (ZrTiO4)-based dielectrics meet all of these requirements (1r < 35; tcf < 2 ppm=8C; tan d < 1:0 1024 ) and have been widely used as dielectric lters, oscillators, and MIC substrates [2]. The microwave dielectric properties of ZrTiO4-based materials are very susceptible to processing conditions. Among these processing variables, the degree of Sn-modication and the cooling rate are known to be the most important parameters [3]. Accordingly, extensive efforts have been made to establish processing-structureproperty relationships [3 9]. It is known that the high-temperature phase of ZrTiO4 polymorphs has the orthorhombic a-PbO2 structure (space
* Corresponding author. Tel.: 82-54-279-2138; fax: 82-54279-2399. E-mail address: [email protected] (H.M. Jang).

group Pbcn) in which two distinct cations, Zr and Ti, randomly distribute over the crystallographically equivalent sites [4]. It undergoes a transition to an incommensurately ordered state at 1125 8C on quasi-static cooling [5,6]. According to the XRD study done by Yamamoto et al. [7], the incommensurately ordered structure can be viewed as a stacking of the 1:1 cation-ordered thin slabs with the interface modulation along the crystallographic a-axis by faulted boundaries. However, the transition to the incommensurate (IC) phase is initiated by the formation of nanoscale cation-ordered clusters in the disordered matrix. Thus, for the study of nano-scale structural features in the beginning stage of the transformation to the IC phase, one has to employ an alternative method because XRD is not able to give information on the short-ranged local structure. The main purpose of the present study is to quantitatively characterize nano-scale structural evolutions associated with the formation of the IC phase from the hightemperature disordered phase by applying the phononconnement model to the analysis of asymmetric Raman line shape. We have particularly paid attention to the morphological evolution of the nano-scale ordered domains with the variation of the cooling rate and the Sn-content that

0038-1098/03/$ - see front matter q 2003 Elsevier Ltd. All rights reserved. doi:10.1016/S0038-1098(03)00463-0

434

Y.K. Kim, H.M. Jang / Solid State Communications 127 (2003) 433437

are considered to be the two most effective parameters for controlling dielectric properties of ZrTiO4-based materials. We have demonstrated that the analysis of the Raman lineshape parameters based on the phonon-connement model is a promising new approach to analyzing the degree of nano-scale atomic positional ordering.

3. Results and discussion 3.1. Mode symmetry of a peak at near 810 cm21 Fig. 1 shows room-temperature Raman spectra of a series of ZTO and ZST polycrystalline ceramics prepared employing various processing conditions. We have presented the Raman band at near 810 cm21 only because this peak exhibits the most pronounced change in the degree of asymmetric line broadening with the variation of processing conditions. Thus, it is useful to clarify the mode symmetry associated with this peak before we will examine the effects of processing conditions on the degree of asymmetric line broadening. The high-temperature cation-disordered ZTO is represented by the a-PbO2 structure with the space group Pbcn and the point group mmm D2h : The normal vibration modes, as obtained by applying the factor group analysis, predict that the high-temperature phase has 18 distinctive Raman-active optical phonon modes [10,11]. These can be summarized using the following non-degenerate irreducible representations:

2. Experimental section ZrTiO4 (ZTO) and Zr12xSnxTiO4 (ZST) polycrystalline ceramic specimens used in this study were prepared employing a high-purity sol gel processing route [3]. The pressed pellets were sintered at 1500 8C for 4 h, and the sintered specimens were cooled with an initial cooling rate of 300 8C/h down to 1250 8C. Various cooling rates were then employed for the subsequent cooling of ZTO for the temperature range between 1250 and 1000 8C. This was done to critically assess the effect of the cooling rate on the degree of short-range ordering associated with the transition to the IC phase. Except for one particular pellet that was directly quenched by immersing it into water at 1250 8C, all other ZTO specimens were then air-quenched immediately after they had reached 1000 8C to isolate the IC phase that was not thermodynamically stable at room temperature. In case of the sintered ZST, the cooling rate between 1250 and 1000 8C was xed at 3 8C/h. Analysis of local composition, as examined using a FESEM/EDS, did not indicate any non-stoichiometric compositional deviation (, 1 at.%) for both interior grain and grain-boundary regions. ICP (inductively coupled plasma) analysis indicated that the total impurity level of ceramic specimens was less than 0.5 at.%. To obtain information on the mode symmetry of Raman peaks, especially for the peak at near 810 cm21, we separately prepared an incommensurately ordered singlecrystal of ZTO. It was grown by slowly cooling (1 8C/h) the solution containing a Li2MoO4 MoO3 ux from 1300 to 900 8C after the mixing process at 1300 8C for 24 h. The single-crystal of ZTO had well dened (001) and (110) planes, and the crystallinity of the single crystal was assessed by examining its XRD u-rocking curve. The measured XRD line was very narrow with the FWHM value less than 0.118, indicating a high degree of the crystallinity. Raman spectra of various ZrTiO4-based ceramics and a single crystal were obtained using a NRS2100 Raman spectrometer (JASCO, Japan) equipped with a triple-grating monochromator and a Coherent Innova 90C Ar-laser operating at 514.5 nm with the power of 300 mW. It has a spectral resolution of 1 cm21. The measurement was performed with a micro-Raman option using a backscattering conguration. For the polarized scattering of the single  crystal ZrTiO4, we employed ZYYZ and ZYXZ scattering geometries, where X; Y; and Z are parallel to [100], [010], and [001] directions, respectively.

GRA 4Ag xx; yy; zz 5B1g xy 4B2g xz 5B3g yz 1

To obtain the mode symmetry associated with the peak at 810 cm21, we have performed polarized Raman scattering on the (001) plane of the incommensurately ordered ZTO   single crystal using both ZYYZ and ZYXZ scattering geometries. As shown in Fig. 2, the peak at near 810 cm21 is  active in the parallel ZYYZ scattering but is not active in  scattering except for a broad peak caused the cross ZYXZ by a polarization leakage. According to Eq. (1) or to the character table for mmm point group, the only normal mode  that is active in the parallel ZYYZ scattering but is inactive

Fig. 1. Raman spectra of ZrTiO4-based dielectrics are shown. (a) Raman spectra of pure ZrTiO4 (ZTO) prepared employing various cooling rates between 1250 and 1000 8C: (i) 1 8C/h, (ii) 5 8C/h, (iii) 10 8C/h, (iv) 50 8C/h, (v) 100 8C/h, (vi) 300 8C/h, and (vii) rapidly quenched. (b) Raman spectra of Zr12xSnxTiO4 (ZST) prepared with various Sn-contents: (i) x 0:01; (ii) x 0:03; (iii) x 0:05; (iv) x 0:07; (v) x 0:12; (vi) x 0:15; (vii) x 0:20; and (viii) x 0:40:

Y.K. Kim, H.M. Jang / Solid State Communications 127 (2003) 433437

435

namely: Iv
BZ

lCq0 0; ql2 dq with v 2 v0 q2 G0 =22

lC0; ql2 exp2q2 L2 =4 where v0 q denotes the phonon dispersion relation, G0 is the intrinsic Raman line width, and L represents the phonon correlation length. As described in Section 1, the incommensurately ordered ZTO can be viewed as a stacking of the 1:1 cation-ordered thin slabs with the faulted boundaries along the a-axis. Thus, the phonon will be one-dimensionally conned within a cation-ordered thin-slab region. However, structural modications either by increasing cooling rate or by increasing Sn-content would gradually decrease the phonon correlation parallel to the faulted boundary, eventually leading to the formation of spherical domains. In this case, the phonon will be three-dimensionally conned within an isolated nanoscaled sphere. Applying this concept to Eq. (2), one can obtain the following equation for the line-shape function: Iv 3.2. Raman line-shape analysis Having identied the mode symmetry corresponding to the peak at 810 cm21, we now examine the effects of processing variables on the line shape of this peak. As shown in Fig. 1, both the degree of the line broadening and the asymmetry in the line shape steadily increase, in addition to the downward shift of the mode frequency, with increasing cooling rate for ZTO or with the degree of Snmodication for ZST. It is well known that the structural characteristics of ordered domains in ZrTiO4-based dielectrics are greatly inuenced by the cooling rate and by the degree of Sn-modication [6,8,12]. Thus, the observed variations of the line-shape parameters are closely related to the structural variations of the cation-ordered domains. To establish quantitative correlations between the Raman line-shape parameters and the structural characteristics of the nano-scale domains, we will apply a concept of phonon-connement [13] to the analysis of spectral line shape. This concept was used to quantitatively explain the broad asymmetric line shapes observed in Si and GaAsbased semiconductors [14,15]. The asymmetry is not expected in a single-crystal having a perfect translational symmetry. In this case, only the Brillouin zone-center modes are Raman-active because of the conservation of crystal momentum. However, the introduction of defects that limit the spatial correlation of phonons then gives rise to a relaxation of the q 0 selection rule. A Gaussian spatial correlation function has been used to account for the qvector relaxation related to nite-size effects and potential uctuations in alloy semiconductors [13 16]. The resulting Raman line shape, Iv, is a superposition of the weighted Lorentzian contributions over the rst Brillouin zone, 1
0

Fig. 2. Polarized Raman spectra of the incommensurately ordered   ZTO single crystal in ZYYZ and ZYXZ scattering geometries.

 in the cross ZYXZ scattering is the Ag normal mode. This clearly indicates that the peak at 810 cm21 is not related to Bg -type normal modes (i.e. B1g ; B2g ; and B3g ) but is caused by the Ag mode that is symmetric to all eight distinctive symmetric operations of mmm point group.

dq v 2 v0 q G0 =2 2
2

f qe2q

2 2

L =4

where f q 1 for a thin-slab while f q 4pq2 for an isolated sphere. In Eq. (3), q is the magnitude of the wavevector in the unit of 2p=a; where a is the lattice constant. Fig. 3 shows a tting example of the Raman line shape using an incommensurately ordered ZTO specimen prepared with the cooling rate of 1 8C/h so that the ordered domains are composed of thin slabs [7]. Thus, f q 1 was used for the tting. To evaluate the intrinsic line-shape parameters in the absence of the connement, we have employed the Raman spectrum of a commensurately ordered Zr5Ti7O24 specimen that does not possess any faulted boundary (i.e. L ! 1) and thus provides us a symmetric reference line for the Ag mode at 810 cm21. It was shown that both the mode frequency and the line width for the Ag normal mode had a good linear correlation with

Fig. 3. Measured Raman spectrum (lled circles) and computed line proles (solid lines) of the incommensurately ordered ZTO. DY is a plot of the tting residuals as a function of the Raman shift.

436

Y.K. Kim, H.M. Jang / Solid State Communications 127 (2003) 433437

the lattice parameters b [12]. The lattice parameter b, as deduced from the XRD pattern of Zr5Ti7O24, was 5.324 A. Then, the corresponding intrinsic mode frequency v0 and the line width G0 ; as estimated from the extrapolated linear relationship for the Ag symmetry [12], were 815 and 29 cm21, respectively. The line width was corrected by the instrumental line width obtained from an Ar -laser line. We then adopted the following one-dimensional periodic relation to evaluate the dispersion relation appeared in Eq. (3):

v0 q A B cosqp

Thus, the intrinsic mode frequency is composed of the two unknown parameters, A and B: To obtain numerical values for A and B; we rst took the estimated value for the average period of the faulted boundary in the incommensurately ordered ZTO (7.46lal or 3.6 nm) as the phonon correlation length for the thin-slab connement. The correlation length of 3.6 nm was estimated from the corresponding XRD superlattice peak. The optimized values of A and B; as obtained from the line-shape tting of Fig. 3 using Eq. (3) with L 3:6 nm and G0 29 cm21 ; were 735 cm21 and 80 cm21, respectively. 3.3. Nano-scale structural evolution Fig. 4 presents the computed variations of the Raman line-shape parameters for two extreme cases (i.e. onedimensional vs. three-dimensional connement). Eqs. (3) and (4) with the optimized values of A and B were used to obtain the two curves. For any particular point on the computed curve, there exists a unique corresponding value of L: In addition to this model computation, we also have plotted the experimental Raman shift and the full width at half-maximum (FWHM) of ZTO and ZST prepared with various processing conditions for the purpose of compari-

son. The Raman spectra given in Fig. 1 were used to extract the experimental line-shape parameters, the Raman shift and the FWHM. The most outstanding difference in the computed lineshape parameters between the two-extreme connements is that the line width increases much more rapidly with the downward shift of the mode frequency for the thin-slab connement. The data points encircled by the dotted line at the lower-right corner of Fig. 4 correspond to ZTO specimens prepared with slow cooling rates (D 1; 5, and 10 8C/h) and ZSTs with low Sn-modication levels (x 0:01 and 0.03). Under these line-narrowing conditions, the experimentally obtained line-shape parameters rather closely follow the theoretical prediction of the thin-slab connement with L between 3 and 4 nm. Beyond this region, however, the line-shape parameters for both ZTO and ZST deviate signicantly from the prediction of the thin-slab model. For the Sn-content higher than 20 at.% (i.e. x $ 0:2), the line-shape parameters for ZST closely follow the prediction of the spherical-connement model. The two data points encircled by the broken solid line at the upper-left corner of Fig. 4 show this trend. The phonon correlation length L in this region had values around 4.0 nm and was not susceptible to the Sn-content. Because the correlation length is essentially independent of the Sn-content for x $ 0:2; the domain size parallel to [100]-direction (or a-axis) is little affected by the Sn-content. Contrary to this, the size of the ordered domain perpendicular to the a-axis must be strongly affected by the Sn-content because of the observed morphological change in the ordered domain from the thin slab to the three-dimensionally conned nano-sphere with increasing Sn-content [8]. As shown in Fig. 4, the line-shape parameters for ZTO do not closely follow the prediction of the spherical connement even under the condition of rapid cooling. Thus, one can conclude that the Sn-modication is a more effective processing parameter than the cooling rate for the stabilization of nano-spherical domain boundaries, which is very important to improving microwave dielectric properties of ZrTiO4-based materials. On the basis of our model computations and analysis of the Raman line-shape parameters, we are now able to delineate the structural evolutions of the nano-scale ordered domains with the variations of the Sn-content and the cooling rate. These are schematically represented in Fig. 5.

4. Conclusions
Fig. 4. Theoretical correlations of the Raman shift for the Ag band with its line width (FWHM) for the two distinctive models of the phonon connement. The line-shape parameters of ZST (open circles) and ZTO (lled triangles) prepared with various processing conditions are also plotted for the purpose of comparison.

The effects of processing variables on the nano-structural characteristics of ZrTiO4-based dielectrics were studied by analyzing the Raman line-shape parameters using the phonon-connement model. The results indicate that the shape of the nano-scale ordered domains undergo a

Y.K. Kim, H.M. Jang / Solid State Communications 127 (2003) 433437

437

References
[1] S. Nomura, Ferroelectrics, 49 (1983) 61. [2] See, for example,W. Wersing, in: B.C.H. Steele (Ed.), Electronic Ceramics, Elsevier, London, 1991, Chapter 4. [3] S.I. Hirano, T. Hayashi, A. Hattori, J. Am. Ceram. Soc. 74 (1991) 1320. [4] P. Bordet, A. McHale, A. Santoro, R.S. Roth, J. Solid State Chem. 64 (1986) 30. [5] A.E. McFale, R.S. Roth, J. Am. Ceram. Soc. 66 (1983) C-18. [6] R. Christoffersen, P.K. Davies, J. Am. Ceram. Soc. 75 (1992) 563. [7] A. Yamamoto, T. Yamada, H. Ikawa, O. Fukunaga, K. Tanaka, F. Marumo, Acta Crystallogr., Sect. C: Cryst. Struct. Commun. 47 (1991) 1588. [8] R. Christoffersen, P.K. Davies, J. Am. Ceram. Soc. 77 (1994) 1441. [9] Y. Park, K.M. Knowles, J. Appl. Phys. 85 (1999) 6434. [10] M.A. Krebs, R.A. Condrate Sr., J. Mater. Sci. Lett. 7 (1988) 1327. [11] F. Azough, R. Freer, J. Petzelt, J. Mater. Sci. 28 (1993) 2273. [12] Y.K. Kim, H.M. Jang, J. Appl. Phys. 89 (2001) 6349. [13] H. Richter, Z.P. Wang, L. Ley, Solid State Commun. 39 (1981) 625. [14] K.K. Tiong, P.M. Amirtharaj, F.H. Pollak, D.E. Aspnes, Appl. Phys. Lett. 44 (1984) 122. [15] P. Parayanthal, F.H. Pollak, Phys. Rev. Lett. 52 (1984) 1822. [16] D. Bersani, P.P. Lottici, X.-Z. Ding, Appl. Phys. Lett. 71 (1998) 73.

Fig. 5. Schematic representation of the effects of the Sn-content and the cooling rate on the nano-scale structural evolution of ZrTiO4based dielectrics.

sequential change from thin-slab form to platelet shape, and nally to spherical shape with increasing Sn-content or cooling rate.

Acknowledgements This work was supported by the KISTEP of Korea through the NRL program.